The watershed transform is a well-established tool for image segmentation. However, watershed segmentation is often
not effective for Synthetic Aperture Radar (SAR) images which are generally corrupted by coherent speckle noise. This
paper presents a novel approach for SAR image segmentation using watershed segmentation algorithm combined with
Otsu. The aim of this study is to improve the generalization of watershed techniques and to construct a well segmentation
of SAR images. Typically, Otsu is used to produce inner mark and external mark. Then we use the mark to modify the
gradient image of the SAR image. Additionally, rather than flooding the gradient image, we use the gradient image
modified by the mark as input to the watershed algorithm. Experimental results demonstrate the superior performance of
the improved watershed-based SAR segmentation.